I want to retrieve the first X and the last Y characters from a string (standard ascii, so no worries about unicode).
I understand that I can do this as seperate actions, i.e. :
FIRST=$(echo foobar | head -c 3)
LAST=$(echo foobar | tail -c 3)
COMBINED= "${FIRST}${LAST}"
But is there a cleaner way to do this ?
I would prefer to use common standard utils (i.e. bash built-ins, sed, awk etc.). At a push, a Perl one-liner is OK, but no Python or anything else.
head + tail two answers, regarding -c switchhead + tail character based (with -c, reducing strings)Under bash, you could
string=foobarbaz
echo ${string::3}${string: -3}
foobaz
But to avoid repetion in case of shorter strings:
if ((${#string}>6));then
echo ${string::3}${string: -3}
else
echo $string
fi
shrinkStr(){
local sep='..' opt OPTIND OPTARG string varname='' paddstr paddchr=' '
local -i maxlen=40 lhlen=15 rhlen padd=0
while getopts 'P:l:m:s:v:p' opt; do
case $opt in
l) lhlen=$OPTARG ;;
m) maxlen=$OPTARG ;;
p) padd=1 ;;
P) paddchr=$OPTARG ;;
s) sep=$OPTARG ;;
v) varname=$OPTARG ;;
*) echo Wrong arg.; return 1 ;;
esac
done
rhlen="maxlen-lhlen-${#sep}"
((rhlen<1)) && { echo bad lengths; return 1;}
shift $((OPTIND-1))
string="$*"
if ((${#string}>maxlen)) ;then
string="${string::lhlen}$sep${string: -rhlen}"
elif ((${#string}<maxlen)) && ((padd));then
printf -v paddstr '%*s' $((maxlen-${#string})) ''
string+=${paddstr// /$paddchr}
fi
if [[ $varname ]] ;then
printf -v "$varname" '%s' "$string"
else
echo "$string"
fi
}
Then
shrinkStr -l 4 -m 10 Hello world!
Hell..rld!
shrinkStr -l 2 -m 10 Hello world!
He..world!
shrinkStr -l 3 -m 10 -s '+++' Hello world!
Hel+++rld!
cnt=1;for str in Généralités Language Théorème Février 'Hello world!';do
shrinkStr -l5 -m11 -vOutstr -pP_ "$str"
printf ' %11d: |%s|\n' $((cnt++)) "$Outstr"
done
1: |Généralités|
2: |Language___|
3: |Théorème___|
4: |Février____|
5: |Hello..rld!|
cnt=1;for str in Généralités Language Théorème Février 'Hello world!';do
shrinkStr -l5 -m10 -vOutstr -pP_ "$str"
printf ' %11d: |%s|\n' $((cnt++)) "$Outstr"
done
1: |Génér..tés|
2: |Language__|
3: |Théorème__|
4: |Février___|
5: |Hello..ld!|
head + tail lines based (without -c, reducing files)By using only one fork to sed.
Here is a little function I wrote for this:
headTail() {
local hln=${1:-10} tln=${2:-10} str;
printf -v str '%*s' $((tln-1)) '';
sed -ne "1,${hln}{p;\$q};$((hln+1)){${str// /\$!N;}};:a;\$!{N;D;ba};p"
}
Usage:
headTail <head lines> <tail lines>
Both argument default are 10.
In practice:
headTail 3 4 < <(seq 1 1000)
1
2
3
997
998
999
1000
Seem correct. Testing border case (where number of line are smaller than requested):
headTail 1 9 < <(seq 1 3)
1
2
3
headTail 9 1 < <(seq 1 3)
1
2
3
Taking more lines: (I will take 100 first and 100 last lines, but print only 2 Top lines, 4 Middle lines and 2 Bottom lines of headTail's output.):
headTail 100 100 < <(seq 1 2000)|sed -ne '1,2s/^/T /p;99,102s/^/M /p;199,$s/^/B /p'
T 1
T 2
M 99
M 100
M 1901
M 1902
B 1999
B 2000
BUG (limit): Don't use this with 0 as argument!
headTail 0 3 < <(seq 1 2000)
1
1998
1999
2000
headTail 3 0 < <(seq 1 2000)
1
2
3
1999
2000
BUG (limit): because of max line length:
headTail 4 32762 <<<Foo\ bar bash: /bin/sed: Argument list too long
For both this to be supported, function would become:
head + tail lines, using one fork to sedheadTail() {
local hln=${1:-10} tln=${2:-10} str sedcmd=''
((hln>0)) && sedcmd+="1,${hln}{p;\$q};"
if ((tln>0)) ;then
printf -v str '%*s' $((tln-1)) ''
sedcmd+="$((hln+1)){${str// /\$!N;}};:a;\$!{N;D;ba};p;"
fi
sed -nf <(echo "$sedcmd")
}
Then
headTail 3 4 < <(seq 1 1000) |xargs
1 2 3 997 998 999 1000
headTail 3 0 < <(seq 1 1000) |xargs
1 2 3
headTail 0 4 < <(seq 1 1000) |xargs
997 998 999 1000
for i in {6..9};do printf " %3d: " $i;headTail 3 4 < <(seq 1 $i) |xargs; done
6: 1 2 3 4 5 6
7: 1 2 3 4 5 6 7
8: 1 2 3 5 6 7 8
9: 1 2 3 6 7 8 9
Stronger test: With bigger numbers: Reading 500'000 first and 500'000 last lines from an input of 3'000'000 lines:
headTail 500000 500000 < <(seq 1 3000000) | sed -ne '499999,500002p'
499999
500000
2500001
2500002
headTail 5000000 5000000 < <(seq 1 30000000) | sed -ne '4999999,5000002p'
4999999
5000000
25000001
25000002
$ perl -E '($s, $x, $y) = @ARGV; substr $s, $x, -$y, ""; say $s' abcdefgh 2 3
abfgh
The four argument variant of substr replaces the given portion of the string with the last argument. Here, we replace from position $x to position -$y (negative numbers count from the end of the string), and use an empty string as replacement, i.e. we remove the middle part.
